pgvector-crystal

Crystal language examples and client usage for pgvector, showing how to work with vector operations in PostgreSQL from Crystal.

• Category: SDKs & Libraries
• Tags: sdk, pgvector, vector-store
• Language / Stack: Crystal, PostgreSQL, pgvector
• License: MIT
• Source: https://github.com/pgvector/pgvector-crystal

Overview

pgvector-crystal provides example code and patterns for using the pgvector extension from Crystal, with support for the crystal-pg PostgreSQL driver. It demonstrates how to perform vector similarity and related operations in PostgreSQL from Crystal-based applications.

Features

• pgvector integration examples for Crystal

  • Demonstrates how to use the pgvector PostgreSQL extension from Crystal code.

• Support for crystal-pg

  • Examples are built around the crystal-pg database library.

• Embeddings workflows

  • Example for using embeddings with OpenAI:
    • How to generate embeddings via OpenAI.
    • How to store and query them with pgvector.

• Binary embeddings

  • Example for using binary embeddings with Cohere:
    • Handling binary embedding representations.
    • Integration with pgvector for storage and search.

• Hybrid search

  • Example with Ollama using Reciprocal Rank Fusion (RRF):
    • Combining multiple ranking signals (e.g., vector and keyword scores).
    • Demonstrates hybrid search patterns with pgvector.

• Sparse search

  • Example of sparse search using text embeddings:
    • Handling sparse vector representations.
    • Querying sparse vectors stored in PostgreSQL.

• Example directory structure

  • examples/openai/example.cr – embeddings with OpenAI.
  • examples/cohere/example.cr – binary embeddings with Cohere.
  • examples/hybrid/example.cr – hybrid search with Ollama (RRF).
  • examples/sparse/example.cr – sparse search with text embeddings.

Getting Started

• Follow the setup instructions for your database library:
  • crystal-pg: primary supported driver (see README in the repository for exact steps).
• Explore the provided Crystal example files in the examples/ directory to learn specific integration patterns.

Pricing

• Not applicable. This is an open-source project released under the MIT license.